SNAP foods typically purchased

README.md

Methods

Based on Table 1 of https://www.fns.usda.gov/sites/default/files/ops/SNAPFoodsTypicallyPurchased-Summary.pdf

absolute_differences.png

analyze.R

library(dplyr)

# load data
dat <- read.csv('snap_report_summary.csv', stringsAsFactors=FALSE)

# annotate SNAP status
snap_cols <- 2:4
non_snap_cols <- 5:7
dat_non_snap <- dat %>% select(1, non_snap_cols)
dat_snap <- dat %>% select(1, snap_cols)
dat_non_snap$household <- "non-SNAP"
dat_snap$household <- "SNAP"

# combine data
names(dat_non_snap)[snap_cols] <- names(dat_snap)[snap_cols]
dat_combined <- rbind(dat_snap, dat_non_snap)

# convert to $
dat_combined$expenditures <- 1e6 * as.numeric(gsub('(\\$|,)', '', as.character(dat_combined$raw)))
dat_combined$percent <- as.numeric(gsub('(\\%)', '', as.character(dat_combined$percent)))

d <- dat_combined # for bevity


# look SNAP vs non-SNAP expenditures
library(tidyr)
d_summary <- select(d, commodity, household, percent) %>%
	group_by(household) %>%
	mutate(id = row_number()) %>%
	spread(household, percent)
d_summary <- d_summary[order(d_summary$SNAP - d_summary$`non-SNAP`, decreasing=TRUE), ]
d_summary$`Broad Category` <- factor(d_summary$commodity, levels=d_summary$commodity)

# does the distribution among categories differ between SNAP and non-SNAP
# households?
# the broad categories used in the summary do not differ
d_chisq <- with(d_summary, chisq.test(`non-SNAP`, SNAP)) # dists are same

library(ggplot2)
png('absolute_differences.png')

ggplot(d_summary) +
	geom_hline(aes(yintercept=0)) + 
	geom_bar(aes(x=`Broad Category`, y=SNAP - `non-SNAP`), stat='identity') +
	ylab("difference in % expenses (SNAP - non-SNAP)") +
	labs(
	title='How food budgets really differ between SNAP and non-SNAP households*',
	caption=paste0("*They don't (p=", round(d_chisq$p.value, 2), " in Chi Sq test).\nBased on Table 1 of\nhttps://www.fns.usda.gov/sites/default/files/ops/SNAPFoodsTypicallyPurchased-Summary.pdf\nSee https://gist.github.com/ashander/6f55b569df0eb933ec1c5d976ff7a849")
	) +
theme(axis.text.x = element_text(angle=90, vjust=0.5))

dev.off()

png('proportional_differences.png')
ggplot(d_summary) +
	geom_hline(aes(yintercept=1)) + 
	geom_bar(aes(x=`Broad Category`, y=SNAP / `non-SNAP`), stat='identity') +
	ylab("proportional % expenses (SNAP / non-SNAP)") +
	labs(
	title='How food budgets really differ between SNAP and non-SNAP households*',
	caption=paste0("*They don't (p=", round(d_chisq$p.value, 2), " in Chi Sq test) but that Baby Food.\nBased on Table 1 of\nhttps://www.fns.usda.gov/sites/default/files/ops/SNAPFoodsTypicallyPurchased-Summary.pdf\nSee https://gist.github.com/ashander/6f55b569df0eb933ec1c5d976ff7a849")
	) +
theme(axis.text.x = element_text(angle=90, vjust=0.5))
dev.off()

process.R

dat <- read.csv('tabula-SNAPFoodsTypicallyPurchased-Summary.csv')
names(dat) <- c("commodity", "rank", "raw", "percent", "rank", "raw", "percent")
write.csv(dat, 'snap_report_summary.csv', row.names=FALSE)

proportional_differences.png

snap_report_summary.csv

tabula-SNAPFoodsTypicallyPurchased-Summary.csv